Iron Triangles and Foreign Exchange [Updated]

My University Politics lecturer Dennis Kavanagh introduced me to the concept of the "Iron Triangle" within US domestic politics. 


Shortly afterwards I was introduced to another Iron Triangle - this time the Project Management "Iron Triangle".
A recent conversation made me re-visit these frameworks within the context of foreign exchange trading.  Let's take the example of a bank that wants to be a price maker.  Clearly the bank will need three things:
  1. Technology to gather market data, process, evaluate, spread, risk manage, distribute, execute, settle and position keeping.
  2. Balance sheet
  3. Client flow
The challenge is that the first part is now quite well understood but some technologists assume that building a world class trading platform will be enough to profitably build market share.  That's a false assumption.  Why?  Simply put:
  • Excellent technology platform + poor client flow + small balance sheet = not profitable
  • Large balance sheet + poor technology platform + poor client flow = not profitable
  • Great client flow + small balance sheet + poor technology platform = not profitable
So, let's continue, what about if you have two good points of the triangle?
  • Excellent technology platform + great client flow + small balance sheet = not profitable
  • Excellent technology platform + large balance sheet + poor client flow = not profitable
  • Poor technology platform + great client flow + large balance sheet  = not profitable
The point here is that each point costs money - great technology increases opex and perhaps capex.  Balance sheet costs money at a compliance and risk adjusted funding rate.  Winning client business requires a sales force and they cost money too...
So how do you win? Hard work.  There's no substitute or quick solution.  Anyone suggesting otherwise is selling snake oil.
Question copied from below:
well then, as with most (if not all) projects, you can't have all 3, what's the best approach?
There's no silver bullet but there is a productive approach. From the overall business perspective it's crucial to understand what the bank wants to achieve and whether that's rational within the market context on a "relative-to-peer" basis. Effectively it's a form of marginal benefit/marginal cost analysis.
So, imagine there are five existing banks in the market and a potential newcomer.  They five banks will each have a mixture of technology/balance sheet/flow and profitability. 
It's going to be possible to evaluate these and provide something approaching a ranking and objective numbers. Of course, there will be a degree of obfuscation and the usual "smoke and mirrors" of business, but some decent analysis and investigation will get some of the way there....
So, below is a set of completely made up numbers:
Technology spend for the FX business on an annualised basis
  1. Bank A $80m/year
  2. Bank B $70m/year
  3. Bank C $60m/year
  4. Bank D $50m/year
  5. Bank E $40m/year
Balance sheet utilisation as measured at 1 day 99% value at risk
  1. Bank A VaR (1 day, 99%) of $20m
  2. Bank B VaR (1 day, 99%) of $10m
  3. Bank C VaR (1 day, 99%) of $30m
  4. Bank D VaR (1 day, 99%) of $55m
  5. Bank E VaR (1 day, 99%) of $40m

Client flow
  1. Bank A daily average $1.5bn
  2. Bank B daily average $1.2bn
  3. Bank C daily average $1.8bn
  4. Bank D daily average $800m
  5. Bank E daily average $750m
Profitability - contribution to bank balance sheet
  1. Bank A $100m/year
  2. Bank B $110m/year
  3. Bank C $75m/year
  4. Bank D $200m/year
  5. Bank E $125m/year
To be sure - there are problems with establishing these numbers - think of the classic case where a bank uses one asset class as a loss leader to gain business from a client firm in another asset class (come trade cheap f/x with us and while you are here, can we interest you in some nice equity IPO allocations?).
The point is, you should be able to conduct a peer evaluation and see how other firms are proceeding in this business line. 
Now, as a new entrant, you will be presented with constraints by the bank for which you are working.  So let's imagine the bank says - you can run a VaR of $tiny.  In which case, you can compare $tiny VaR to peers and conduct analysis - consider the idea that for each marginal dollar of VaR you may need to increase technology spend and bring in client flow to utilise that extra dollar.
As you can see - what we have here is a constrained optimisation problem that is familiar to any Economist, Statistician or "Data Scientist" in the latest buzzword.
So, in summary:
  • Establish constraints on VaR
  • Conduct peer analysis
  • Estimate relative costs of increasing VaR and client flow
  • Estimate costs of build out of technology stack sufficient to service the client flow at that level of VaR
  • Analyse the optionality embedded in the technology stack - will it be creaking at that scale of business or can it scale up? Would it make sense to move to a stack that scales better?
  • Analyse distribution channels available for business.  Should the firm continue with old fashioned channels such as telephone and SDP or look to move to an API (FIX, ITCH/OUCH) based distribution model? Calculate the marginal costs of each distribution channel - if providing FIX brings 90% of flow at 5% of the cost versus voice bringing 10% of flow at 95% of the cost  - why offer voice? If an SDP will cost $50m but bring minimal profit due to high cost of servicing the client - why offer and SDP?
  • Examine costs holistically.  Don't just view spending on technology as a sunk cost - view it as the cost of doing business. 
  • "We've always done it this way" is a good way to spend $$$$ for no return

In order to better show this I have transferred the data above into R and when I get some more time I will put together an optimisation model and dump it onto Github.  But that will take a while and I am a bit busy at the moment...


R plot of the data - Client Flow


R plot of the data - Technology Spend

R plot of the data - Value At Risk


R plot of the data - Profit 

Comments